The Future of Work

Automation, digitization, and trade have dramatically reshaped how firms organize, operate, and compete over the past couple of decades, and these changes have implications for the future of work. Academic research is only just beginning to understand how these changes affect workers, incentives to create new firms and innovations, and impact local and regional economies.


Automation, digitization, and trade have dramatically reshaped how firms organize, operate, and compete over the past couple of decades. These changes, in turn, have had dramatic consequences for labor. Academic research has an important role to play in understanding these trends. Recent research in this area contemplates their continued effects on labor and, ultimately, what the future of work will look like. While popular media often describe dystopian futures in which robots have all the jobs, a consideration of episodes of automation in the past suggests that many jobs will be replaced, but many more new jobs will be created (Autor 2015).

Important questions remain, however, about how automation, digitization, and trade will affect households, firms, and geographies, including the potential for increased inequality and differential effects by race, gender, and socioeconomic background. Policymakers will likely grapple with these important issues in the coming years, and academic research will be an essential input to help inform both policymakers and practitioners.


Both policymakers and academics are concerned with slowing economic growth and slowing rates of new firm formation. There is also evidence of increasing dispersion in productivity, which suggests that some firms are able to take advantage of new technologies and globalization while others are not (Decker et al 2014, 2016). Advances in automation, including increased use of robotics and artificial intelligence, as well as more outsourcing of parts of the supply chain have important implications for the future of work. Technology may also be allowing firms to connect with workers, suppliers, and customers in new ways, allowing for new work arrangements, including working remotely, working on virtual teams, working on-call, or working as independent contractors as part of the gig-economy.

Globalization sends jobs abroad through offshoring. Acemoglu, Gancia and Zilibotti (2014) consider the impact of offshoring on the U.S. labor market and income inequality, arguing that offshoring low-skill jobs decreases the cost of low-skill products and therefore increases the relative price of skill-intensive products. This relative price effect would induce more innovation in skill-intensive sectors. Lower costs, however, also increase profits, creating a market size effect through which innovation incentives are increased for low-skill goods. Whether the relative price effect or the market size effect dominates depends on the level of offshoring. The authors argue that at its early stages (when offshoring is low), the price effect dominates and skill-biased technical change occurs, and income inequality worsens. As offshoring becomes more common, the market size effect starts to dominate and innovation shifts to low-skill sectors, and the skill premium and income inequality are reduced. The impact of offshoring, then, depends on the level of offshoring, and the distributional implications may be quite different over time.  

Throughout history, there has been concern that automation, including mechanization, computing, and more recently robotics, would kill jobs and generate irreversible damage to the labor market. For example, Keynes (1930) described technological unemployment as "unemployment due to our discovery of means of economising the use of labour outrunning the pace at which we can find new uses for labour." Similarly, Leontief (1983), observing the dramatic improvements in the processing power of computer chips, worried that people would be replaced by machines, just like horses were made obsolete by the invention of internal combustion engines.

Despite this pessimism, there is ample evidence that automation fosters productivity growth (e.g., Bloom, Sadun, and Van Reenen 2012). Economic research shows that, in the past, automation has often substituted for human labor in the short term, but has led to the creation of complementary jobs in the long term (Autor 2015). Automation appears to have had different effects by occupation. For example, historically, it appears that middle-skill jobs have been displaced by automation, leading to labor market polarization (Autor and Dorn, 2013), though there is some evidence that labor market polarization is declining (Schmitt, Shierholz, and Mishel, 2013),.

Projecting into the future is more difficult, but work by Frey and Osborne (2013) and McKinsey Global Institute (2015) suggests that occupations with many routine functions face a higher probability of automation. CEA (2016) uses the Frey and Osborne data to show that the jobs most likely to be automated are the lower-wage jobs.  From a policy perspective, these potential changes make it important to help workers acquire relevant new skills. Recent policy articles have therefore argued for more investment in job training and re-training programs, and expansion of the Earned Income Tax Credit (McAfee and Brynjolfsson 2016, Seamans 2017b). More dramatic policy proposals, that need more research before being considered, include a universal basic income, which seeks to replace existing safety net programs with a single, unconditional cash transfer.

Robots can be thought of as a specific form of automation, and thus it is likely that robots also foster productivity growth. Recent academic research, using national level data on worldwide robot shipments, suggests that robots may have been responsible for about one-tenth of the increases in gross domestic product (GDP) between 1993 and 2007 (Graetz and Michaels 2015).  According to the 2016 Economic Report of the President, worldwide demand for robots has nearly doubled between 2010 and 2014 (CEA 2016, Figure 5-11), and the number and share of robot-related patents have also increased (CEA 2016, Figure 5-14). Thus, robots may now be contributing even more to GDP growth than in the past.

While there is a tendency to anthropomorphize robots, or to think of them as robot arms in a manufacturing setting, there are, in fact, many types of robots, including virtual or software robots. For example, people often refer to Apple's voice-activated Siri and Amazon's Alexa as robots. More broadly, there are a whole range of types of robots enabled by artificial intelligence (AI). There has been much research in the computer science discipline on AI, but economists have only just started to investigate the ways that AI may impact the workforce (e.g., Agrawal, Gans, Goldfarb 2016).

Even as robots may be contributing to GDP growth at a national level, we lack an understanding about how robots may be contributing to new firm formation and attendant job growth (or loss). As with prior episodes of automation, robots may serve to complement or substitute for labor (Brynjolfsson and McAfee, 2014), and these effects may vary in the short- and long-run (Acemoglu and Restrepo 2016).

Acemoglu and Restrepo (2016) point to the race between automation by machines and new complex tasks by humans. They argue that there is a continuous process in which tasks previously performed by labor become mechanized and new employment opportunities in more complex tasks are being produced. Acemoglu and Restrepo argue that the main difference between horses and humans is that humans have comparative advantage in complex tasks and horses do not. For example, the Second Industrial Revolution led to the replacement of the stagecoach by railroad and of sailboats by steamboats. The same change produced new tasks for humans, such as engineers, repairers, conductors, and modern managers (Mokyr 1990). Acemoglu and Restrepo (2016) show that since 1980, employment growth has been highest in occupations with more new job titles, suggesting rapid changes in the nature of work for these sectors of the economy.

The effect of robots on jobs and productivity likely vary across industry, occupation, firm size, and region of the country. Robots may also have implications for entrepreneurial firms, which may lack knowledge of how to best integrate robots with a workforce and may face financing constraints that make it harder for them to adopt capital-intensive technologies in cases where physical robots are needed. These firms may also face other barriers in terms of access to relevant "big data" (see section below) in cases where data is needed to train AI systems.

Thus, while robotics may lead to overall GDP growth, it is not clear whether this growth comes at the expense of jobs in the short-term, or at the expense of new firm formation. There is a need for more research in this area. In particular, there is a lack of systematic data on the use of robots or other forms of automation (Seamans 2017a). Researchers and policymakers would be well-served to collect this data and put it in the public domain.

Changing (offline) workplace
One of the most striking trends in the last decade is the rise of alternative work arrangements—defined by Katz and Krueger (2016) as "temporary help agency workers, on-call workers, contract workers, and independent contractors or freelancers." These arrangements are increasingly common. Moreover, Katz and Krueger (2016) find that alternative work arrangements have been responsible for over 90 percent of the total growth in jobs in the U.S. economy since 2005.

In a study of Uber driver survey results, Hall and Krueger (2015) find that this form of employment has become more popular, in part, as a result of worker desire for flexibility. However, field experiments about call center hiring by Mas and Pallais (2016) suggest that "[T]he great majority of workers are not willing to pay for flexible scheduling relative to a traditional schedule."

Bloom et al. (2014) reports the results of a field experiment in working from home ("WFH"). They show there is strong demand from workers for WFH policies in the workplace. Moreover, WFH increased worker productivity and increased retention and worker satisfaction, but it decreased promotion rates for workers who opted in.

Another recent trend in the labor force has been companies' use of "big data" and "workforce analytics." Although these methods are technologically new, they connect to an earlier formal theoretic literature about improved monitoring, screening, and coordination in contracts. In the hiring and screening setting, Cowgill (2017), Hoffman, Kahn, and Li (2016) and Horton (2016) show evidence that the adoption of rules in decision-making improves hiring outcomes compared to the application of human discretion. There has been less work about the monitoring and coordination benefits of "workforce analytics."

Gig economy/ sharing economy: The changing online workplace
Work has also changed in recent decades because of the prevalence of online labor markets such as Uber, TaskRabbit, Upwork, Mechanical Turk, and others. Although current estimates by Katz and Krueger suggest that this form of work is still very rare (0.5 percent of workers), it is rapidly growing. Research by the JP Morgan Institute estimates that, cumulatively, less than 1 percent of U.S. consumers have ever relied on online platforms for work (JP Morgan Institute 2016). The "online" aspect of these markets is the matching of buyers and sellers. The actual work—such as driving a car or mowing a lawn—often is performed offline.

These matching mechanisms offer several benefits to workers and employers compared to prior markets. The first benefit is an improved contracting environment. In online labor markets, contracts are more standardized and better enforced than in their offline counterparts. For example, Uber and Lyft have clear expectations about cost, cleanliness, timeliness, and the behavior of riders and drivers. Absent the standardization brought by these companies, drivers and riders would have to negotiate these expectations separately for each transaction.

Online contracts are also better for reputational and enforcement reasons. Prior to Uber and Lyft, taxi riders or drivers may have been tempted to cheat the other party. A driver, for example, could leave the rider at an undesired destination or a rider could refuse to pay the driver the full amount. Platforms that intermediate these transactions can investigate reports of bad behavior and adjudicate disputes. They can also pool reports about worker (and client) quality in a way that disciplines behavior ex-ante.

In addition to better contracting, "gig economy" marketplaces also offer participants lower search costs. Fradkin (2014) uses results to show that lower search costs substantially boost exchange rates on Airbnb. Cramer and Krueger (2015) show that Uber drivers are more efficient than standard taxi drivers (they spend more time driving and drive a higher number of miles with passengers). They attribute this efficiency benefit to the online marketplace's superior search and matching technology, compared with the offline system of search and matching.

Although lower search costs increase efficiency, social welfare, and match quality, lower search may also affect the distribution of utility in adverse ways. Some parties may actually be harmed by lower search costs. Classic models of search by Diamond (1971) show that even small search costs may permit monopolist-like pricing. Some sellers—particularly those with strong brands —may actually desire higher search costs for the rest of the market. Lower search costs may permit easier comparison between alternatives and thus encourage price competition. Ride-sharing services portray drivers as (essentially) commoditized, undifferentiated substitutes. The resulting competition in prices would decrease worker profits.

Although lower search costs could possibly harm undifferentiated workers, these losses may be offset by an increase in demand. The benefits of online marketplaces may attract new employers into the marketplace who bid up the wages of workers even in a commoditized marketplace. Using a difference-in-difference approach to study the effect of Uber's entrance into a geographic market on income for taxi and limousine drivers, Cramer (2016) finds no effects.

For workers, one of the main benefits of the "gig" economy is the flexibility around the timing and conditions of work (Hall and Krueger, 2015). Much of the platforms-based work, however, is on a contract basis, with few or no guarantees of benefits and regular income. Why are workers willing to pursue this form of employment without benefits? One reason suggested by Hall and Krueger (2015) is that "gig" work is often a second source of income. The primary employer may provide benefits, and there is no need for additional benefits.

In addition, some scholars (Warshawsky, 2016) have argued that the U.S. tax code creates incentives to pay workers in benefits they don't desire or utilize. U.S. law requires certain benefits subsidies to be equal for all workers, resulting in one-size-fits-all benefits policies for entire firms. The result is a mismatch between the benefits workers desire and those offered by prospective employers. This mismatch may explain why many workers appear comfortable forgoing traditional employment for alternative arrangements. Alternative arrangements are typically paid in cash, and benefits can be purchased according to individual preferences.

Future Research

As highlighted in the paragraphs above, much of the research in this area is nascent. There are a wide range of research questions for scholars to pursue, including:

  1. Which types of firms are more likely to invest in robots and automation (by industry, age, size, geography)?
  2. Should robots or other forms of automation be taxed (or subsidized)?
  3. What is the impact of robots and automation on productivity? To what extent are robots substituting or complementing for labor (by industry, age, size, and geography)?
  4. What are the distributional consequences of automation? What is the impact on inequality and welfare? What is the aggregate welfare implication? What are the right policies to make this process more inclusive? And how best to bring in people left behind?
  5. What are the relevant education and training policies to minimize the cost of transition?
  6. Do existing safety nets (e.g., unemployment insurance, which varies by state) have enough financial resources to address any increase in unemployment due to robots and automation?
  7. How does the downstream (or upstream) market structure affect a firm's incentives to invest in robots and automation? How does the use of robots in a focal market affect new firm entry rates in its market (or upstream or downstream markets)?
  8. What kinds of management practices are correlated with the adoption of robots and automation?
  9. What is more important for online labor markets: better contracts or better search?
  10. What are the relevant competition policies to create new jobs?
  11. How do "people analytics" practices improve outcomes besides hiring?

Data sources

As noted above, there is a need for more systematic data measuring how people work and how firms use robotics and automation. Some existing data sources that are publicly available or available for a fee include:


Acemoglu, Daron and Pascual Restrepo. 2016. "The Race Between Machine and Man: Implications of Technology for Growth, Factor Shares and Employment," MIT working paper.

Agrawal, A., C. Catalini, and A. Goldfarb. 2013. "Some Simple Economics of Crowdfunding," in Innovation Policy and the Economy, Volume 14, Josh Lerner and Scott Stern editors. NBER, University of Chicago Press.

Agrawal, A., J. Gans, and A. Goldfarb, 2016. "The Simple Economics of Machine Intelligence." Harvard Business Review, 17.

Autor, David H. 2015. "Why Are There Still So Many Jobs? The History and Future of Workplace Automation." Journal of Economic Perspectives 29(3): 3–30.

Autor, David H. and David Dorn. 2013. "The Growth of Low-Skill Service Jobs and the Polarization of the US Labor Market." American Economic Review 103(5): 1553–1597.

Bloom, Nicholas, Raffaella Sadun, and John Van Reenen. 2012. "Americans Do I.T. Better: US Multinationals and the Productivity Miracle." American Economic Review 102(1): 167–201.

Brynjolfsson, Erik and Andrew McAfee. 2014. The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies, W. W. Norton & Company.

Burtch, Gordon, Seth Carnahan, and Brad N. Greenwood. 2016. "Can You Gig it? An Empirical Examination of the Gig-Economy and Entrepreneurial Activity." U Michigan working paper.

Council of Economic Advisers (CEA). 2016. Economic Report of the President.

Cramer, Judd and Alan B. Krueger. 2016. "Disruptive Change in the Taxi Business: The Case of Uber." American Economic Review, 106(5): 177-82.

Decker, Ryan, John Haltiwanger, Ron Jarmin, and Javier Miranda. 2014. "The role of entrepreneurship in US job creation and economic dynamism." The Journal of Economic Perspectives 28 (3): 3-24.

Decker, R. A., J. Haltiwanger, R. S. Jarmin, & J. Miranda. 2016. "Declining business dynamism: Implications for productivity?" Brookings Working Paper.

Frey, Carl Benedikt and Michael A. Osborne. 2013. "The Future of Employment: How Susceptible Are Jobs to Computerisation?" Oxford Martin Programme on the Impacts of Future Technology.

Graetz, Georg and Guy Michaels. 2015. "Robots at Work." Centre for Economic Performance Discussion Paper No. 1335.

Hall, Jonathan V. and Krueger, Alan B. 2015. "An Analysis of the Labor Market for Uber’s Driver-Partners in the United States" Working Papers (Princeton University. Industrial Relations Section) ; 587

JPMorgan Chase Institute. 2016. "Paychecks, Paydays, and the Online Platform Economy" February 2016.

Katz, L. F. and A. B. Krueger. 2016. The Rise and Nature of Alternative Work Arrangements in the United States, 1995-2015.

Keynes, John Maynard. 1930. "Economic Possibilities for our Grandchildren." In Essays in Persuasion. London: Macmillan and Co., Ltd.

Leontief, Wassily. 1983. "National Perspective: The Definition of Problems and Opportunities." National Research Council. The Long-Term Impact of Technology on Employment and Unemployment. Washington, DC: The National Academies Press. doi: 10.17226/19470.

McAfee, A. and E. Brynjolfsson. 2016. "Human Work in the Robotic Future" Foreign Affairs July/August 2016 Issue.

McKinsey Global Institute. 2015. "Four Fundamentals of Workplace Automation." McKinsey Quarterly, November 2015.

Mollick, E. R., and R. Nanda. 2014. "Wisdom or Madness? Comparing Crowds with Expert Evaluation in Funding the Arts," HBS working paper.

Schmitt, John, Heidi Schierholz, and Lawrence Mishel. 2013. "Don't Blame the Robots. Assessing the Job Polarization Explanation of Growing Wage Inequality." Economic Policy Institute.

Seamans R. 2017a. " We Won't Even Know If A Robot Takes Your Job" January 11, 2017.

Seamans R. 2017b. "No, Robots Should Not Be Taxed" March 3, 2017.

Sundararajan, A. 2016. The Sharing Economy: The End of Employment and the Rise of Crowd-Based Capitalism. MIT Press.